Systems | Information | Learning | Optimization
 

Privacy-assurances in multiple data-aggregation transactions

Aggregating data from a large number of users is integral part of the business model of many companies. As a result, people are increasingly concerned about privacy of their data. In this talk, I will present a few privacy-preserving algorithms for aggregating data from a large number of users at a third-party application. We consider aggregation through two commonly
used functions, weighted sum and maximum. Unlike traditional secure multi-party computations, our algorithms are designed for scenarios where the users do not know each other and they are not eager to exchange information with each other. In fact, a key aspect of these algorithms is that they do not require any direct communication between two users.
Another important aspect of these algorithms is that the privacy assurances are provided even when the users are involved in multiple transactions with the same third-party application. The second aspect is a major issue that has not been widely addressed in literature. Finally, the algorithms do not impose significant computational and communication burden on any user.
In particular, the computational and communication complexity required of an user does not grow with number of users.
March 12 @ 12:30
12:30 pm (1h)

Discovery Building, Orchard View Room

Parmesh Ramanathan